Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised controlled trials

School of Social and Community Medicine, University of Bristol, Bristol BS8 2PS, UK.
BMJ (online) (Impact Factor: 17.45). 01/2011; 343:d4002.
Source: PubMed
Download full-text


Available from: Sue Duval, Jul 20, 2015
  • Source
    • "The first possible interpretation is that this shows publication bias. However, another possibility is that given the scalability of plate size, the variation in standard errors may represent different strength manipulations of plate size (Sterne et al. 2011). That a large change in plate size might result in a larger effect is something we will capture later by calculating the plate size elasticity of consumption. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The literature on whether varying plate size has an effect on consumption is mixed and contradictory. This meta-analysis of 56 studies from 20 papers shows that varying the size of the container holding food (e.g., plate or bowl) has a substantial effect on amount self-served and/or consumed (Cohen’s d=.43). More generally, we found a doubling of plate size increased the amount self-served or amount consumed by 41%. Our analysis resolves the various contradictions of past reviews: we found that the plate size-effect had a substantial effect on amount served (d=.51), and on amount consumed when the portion was self-served (d=.70) or manipulated along with (confounded with) plate size (d=.48). However, plate size had no effect on amount consumed when the portion size was held constant (d=.03). Overall, plate size had a stronger effect when participants were unaware that they were participating in a food study (d=.76).
    Full-text · Article · Jan 2016
    • "Since random effect models do not provide a quantitative measure of heterogeneity, we assessed study heterogeneity through visual inspection of funnel plots (Egger et al., 1997). Studies were excluded from the analysis in a stepwise manner when they fell outside the range where 95% of studies would be expected in the absence of bias and/or heterogeneity (Sterne et al., 2011). Post-hoc comparisons between subgroup effects were assessed using independent samples t-tests and potential moderator effects were assessed by meta-regression analysis. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Non-pharmacological therapies, such as physical activity interventions, are an appealing alternative or add-on to current pharmacological treatment of cognitive symptoms in patients with dementia. In this meta-analysis, we investigated the effect of physical activity interventions on cognitive function in dementia patients, by synthesizing data from 802 patients included in 18 randomized control trials that applied a physical activity intervention with cognitive function as an outcome measure. Post-intervention standardized mean difference (SMD) scores were computed for each study, and combined into pooled effect sizes using random effects meta-analysis. The primary analysis yielded a positive overall effect of physical activity interventions on cognitive function (SMD[95% confidence interval] = 0.42[0.23;0.62], p < .01). Secondary analyses revealed that physical activity interventions were equally beneficial in patients with Alzheimer's disease (AD, SMD = 0.38[0.09;0.66], p < .01) and in patients with AD or a non-AD dementia diagnosis (SMD = 0.47[0.14;0.80], p < .01). Combined (i.e. aerobic and non-aerobic) exercise interventions (SMD = 0.59[0.32;0.86], p < .01) and aerobic-only exercise interventions (SMD = 0.41[0.05;0.76], p < .05) had a positive effect on cognition, while this association was absent for non-aerobic exercise interventions (SMD = -0.10[-0.38;0.19], p = .51). Finally, we found that interventions offered at both high frequency (SMD = 0.33[0.03;0.63], p < .05) and at low frequency (SMD = 0.64[0.39;0.89], p < .01) had a positive effect on cognitive function. This meta-analysis suggests that physical activity interventions positively influence cognitive function in patients with dementia. This beneficial effect was independent of the clinical diagnosis and the frequency of the intervention, and was driven by interventions that included aerobic exercise.
    No preview · Article · Nov 2015 · Ageing research reviews
  • Source
    • "Sensitivity analysis was performed by assessing the contribution of individual studies to the summary effect estimate by excluding each study, one at a time, and computing meta-analysis estimates for the remaining studies . With only eight studies the power of testing for publication bias might be too low to distinguish chance from real asymmetry (Sterne et al., 2011), therefore funnel plot and Egger's regression analysis were omitted. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Birth weight is extensively investigated as an outcome of interacting with greenery in mothers’ living environment, because it is one of the major causes for neonatal and infant mortality, as well as a correlate of some adverse effects in childhood and beyond. Conversely, in modern urban-ecological systems the access to greenery is limited. The aim of this study was to answer the question whether green spaces and generally greenery in the living environment of pregnant women are associated with the birth weight of their infants and what the direction of that effect is. MEDLINE, EMBASE and the Internet were searched for relevant publications in English and Spanish. Eight studies were identified and included in the analyses (total n = 214 940). We report quality effects meta-analyses based on correlation and standardized regression coefficients as estimates of effect size. Neighbourhood greenness within 100-m buffer was weakly but positively associated with birth weight. The pooled correlation coefficient was 0.049 (95% CI: 0.039, 0.059) and the pooled standardized regression coefficient was 0.001 (95% CI: -0.001, 0.003). There was, however, considerable heterogeneity between the studies. Using more sensitive measures for greenness and taking into account green space functionality and quality, adjusting for environmental exposures, and assessing individual attitudes towards nature, might yield clearer picture, higher statistical power and more precise results in future research. Our findings endorse the emphasis put on urban forestry and landscape management as closely related to public health and propose a more naturalistic, humanitarian and person-centered approach in future studies.
    Full-text · Article · Sep 2014 · Urban Forestry & Urban Greening
Show more